search

ZongxuPan / DrBox-v2-tensorflow

The tensorflow implementation of DrBox-v2 which is an improved detector with rotatable boxes for target detection in remote sensing images
Apache License 2.0
47 stars 21 forks source link

Problems with my own dataset #4

Open niuxiansen2 opened 5 years ago

niuxiansen2 commented 5 years ago

hello @ZongxuPan

Is the size of the input image fixed? Why do I encounter the following problems with my own data set?

Traceback (most recent call last): File "Drbox.py", line 469, in net.train() File "Drbox.py", line 284, in train input_im[k] = im.reshape(IM_HEIGHT, IM_WIDTH, IM_CDIM).astype('float32') ValueError: cannot reshape array of size 496500 into shape (300,300,3)

ZongxuPan commented 5 years ago

hello @ZongxuPan

Is the size of the input image fixed? Why do I encounter the following problems with my own data set?

Traceback (most recent call last): File "Drbox.py", line 469, in net.train() File "Drbox.py", line 284, in train input_im[k] = im.reshape(IM_HEIGHT, IM_WIDTH, IM_CDIM).astype('float32') ValueError: cannot reshape array of size 496500 into shape (300,300,3)

Please set IM_HEIGHT and IM_WIDTH as the desired values.

ZongxuPan commented 5 years ago

hello @ZongxuPan

Is the size of the input image fixed? Why do I encounter the following problems with my own data set?

Traceback (most recent call last): File "Drbox.py", line 469, in net.train() File "Drbox.py", line 284, in train input_im[k] = im.reshape(IM_HEIGHT, IM_WIDTH, IM_CDIM).astype('float32') ValueError: cannot reshape array of size 496500 into shape (300,300,3)

Besides, FEA_HEIGHT4, FEA_WIDTH4, FEA_HEIGHT3, FEA_WIDTH3 are needed to be reset. FEA_HEIGHT4 and FEA_WIDTH4 are the size of the feature map of conv4_3. FEA_HEIGHT3 and FEA_WIDTH3 are the size of the feature map of conv3_3.

niuxiansen2 commented 5 years ago

hello @ZongxuPan Is the size of the input image fixed? Why do I encounter the following problems with my own data set? Traceback (most recent call last): File "Drbox.py", line 469, in net.train() File "Drbox.py", line 284, in train input_im[k] = im.reshape(IM_HEIGHT, IM_WIDTH, IM_CDIM).astype('float32') ValueError: cannot reshape array of size 496500 into shape (300,300,3)

Besides, FEA_HEIGHT4, FEA_WIDTH4, FEA_HEIGHT3, FEA_WIDTH3 are needed to be reset. FEA_HEIGHT4 and FEA_WIDTH4 are the size of the feature map of conv4_3. FEA_HEIGHT3 and FEA_WIDTH3 are the size of the feature map of conv3_3.

Okay, I understand，Thank you very much!

niuxiansen2 commented 5 years ago

hello @ZongxuPan Is the size of the input image fixed? Why do I encounter the following problems with my own data set? Traceback (most recent call last): File "Drbox.py", line 469, in net.train() File "Drbox.py", line 284, in train input_im[k] = im.reshape(IM_HEIGHT, IM_WIDTH, IM_CDIM).astype('float32') ValueError: cannot reshape array of size 496500 into shape (300,300,3)

Besides, FEA_HEIGHT4, FEA_WIDTH4, FEA_HEIGHT3, FEA_WIDTH3 are needed to be reset. FEA_HEIGHT4 and FEA_WIDTH4 are the size of the feature map of conv4_3. FEA_HEIGHT3 and FEA_WIDTH3 are the size of the feature map of conv3_3.

When I train with my own data set, I don't know how to set the values of PRIOR_HEIGHTS and PRIOR_WEIGHTS. What is your method?